To spot a big improvement, say from 5% to 7.5% of visitors becoming leads, you need roughly 1,500 visitors on each version, or about 3,000 in total. Smaller improvements need far more.
That is the first decision for a low-traffic campaign: can you bring enough people to both versions while the offer and audience stay reasonably consistent? Work out the likely waiting time before splitting your traffic.
This guide shows the visitor numbers behind that decision, what to measure and what to improve when a useful test would take too long.
How many visitors and leads do you need
Start with the share of visitors who currently become leads and the improvement you want to measure. The table shows how the visitor requirement changes when you aim to detect a bigger or smaller improvement.
| Assumed rate change | Relative lift | Visitors per version | Total visitors | Expected leads A and B |
|---|---|---|---|---|
| 2% to 3% | 50% | About 3,900 | About 7,800 | 78 and 117 |
| 5% to 7.5% | 50% | About 1,500 | About 3,000 | 75 and about 113 |
| 5% to 6% | 20% | About 8,200 | About 16,400 | 410 and 492 |
Planning note: Calculated examples, not client results or guaranteed outcomes. Assumptions: two-sided 5% significance, 80% power, equal allocation and a fixed-sample normal approximation (pooled null variance, unpooled alternative variance; statsmodels method). Visitor counts round up to the next 100 per version. Expected leads are averages at the assumed rates, not a minimum-conversion threshold.
Moving from 5% to 7.5% means a 50% relative improvement. Moving from 5% to 6% means a 20% relative improvement. The smaller change needs about 16,400 visitors across the two versions, compared with about 3,000 for the larger change.
Choose an improvement that would matter to the business and that your proposed change could plausibly produce. A calculator can help you plan the test; the customer problem should determine what you test.
Work out how long you would need to wait
Divide the total visitors needed by the number of eligible visitors you receive each day. For the 3,000-visitor example, 100 eligible visitors a day means about 30 days. At 25 a day, it means about 120 days.
Ask whether the campaign can stay consistent for that long. If the offer expires soon, the audience is changing or a busy season is about to end, a long test may answer a question that no longer matters by the time it finishes.
Plan around the working patterns that affect your customers, including differences between weekdays and weekends. Optimizely’s guidance on sample size and experiment duration explains why visitor numbers and the time covered both matter.
Check that you are counting useful leads
For a lead-generation page, count valid inquiries. Decide how spam and duplicate submissions will be handled, and use the same rules for both versions. Count eligible people entering the test rather than adding another visitor every time someone reloads the page.
Submit the form yourself before starting. Check that the inquiry reaches the right person and that the reporting records it once. Ask whoever manages the website to use Google’s analytics verification tools to check event receipt as well.
Also decide what makes a lead useful. More submissions can create extra work if they come from people who need a different service or are outside your service area. Review the quality of the inquiries alongside the total.
Give each version one clear job
Start with a specific question. For example, will more suitable prospects contact you if the headline clearly explains the service and what happens after they submit the form? Compare the existing page with a version built around that idea.
Run the two versions at the same time. Use a testing tool to assign visitors randomly, and keep each person on the same version when they return. Keep the offer, traffic rules and lead definition consistent across the comparison.
Write down the planned visitor target and stopping rule before launch, using your testing tool’s guidance. Check for broken forms and tracking failures during the test, and handle those faults promptly. For a test planned around a set visitor target, wait for the planned finish before choosing a winner.
At the end, review the difference, the uncertainty reported by the tool and the quality of the leads. If the result is too close to call, record it as inconclusive and use what you learned to choose the next question.
What to change when traffic is too low
A long waiting time is a reason to change your approach. Start with problems you can observe directly instead of leaving a broken or confusing page in place.
- Fix forms that fail, missing inquiry notifications and duplicate tracking events.
- Match the page headline and offer to the ad or link that brought the visitor there.
- Make the next step clear, including what information to provide and what happens after the form is sent.
- Check the page on a phone for hard-to-read copy, awkward fields and navigation problems.
- Use recent customer questions and lead-quality feedback to identify information that serious prospects need.
Make a focused improvement, record the date and watch what happens to inquiries and their quality. Keep a note of changes to ads, offers and traffic so you can interpret the result. Use these observations to plan a later randomized test when enough traffic is available.
Button clicks and form starts can help you find where visitors hesitate. Keep valid inquiries as the main measure when the business goal is to generate leads.
Use a short plan before you launch
A short written plan helps you decide whether testing is practical and keeps everyone working toward the same customer action.
| Planning field | What to record |
|---|---|
| Useful lead | The customer action being counted and how spam or duplicate submissions are handled |
| Visitors included | Who can enter the test and how returning visitors are counted |
| Current lead rate | The share of visitors who become valid leads under the same counting rules |
| Improvement worth testing | The improvement that would justify a change and why it is plausible |
| Visitors needed | The target for each version and the testing tool’s planning settings |
| Waiting time | How long the traffic target will take and whether the campaign can stay consistent |
| Finish and review | When to stop, how to monitor technical faults and how to review lead quality |
If the visitor target and waiting time fit the campaign, run the comparison. If they do not, fix the clearest obstacles, keep monitoring lead quality and return to testing when the numbers make the decision useful.
For help connecting the page, lead tracking and testing plan, explore Around29’s landing page design and testing.